Thermal fluids are critical to technologies such as electric vehicles, data centers, and power electronics, where rising power densities demand fluids with increasingly challenging performance requirements. Designing these fluids is a complex multi-objective optimization problem across vast molecular and formulation spaces, making traditional experimental and computational approaches inefficient.

This talk presents a closed-loop thermal fluid discovery framework that integrates AI, automation, and molecular simulation. A self-driving laboratory (SDL) combines high-throughput experimentation with machine learning to autonomously explore formulation spaces. Using Feature-Adaptive Mixture Bayesian Optimization (FAMBO), the platform efficiently identifies high-performance formulations while substantially reducing experimental effort. Molecular dynamics simulations and representation learning further reveal the molecular origins of nonlinear mixture behavior, enabling predictive and scientifically interpretable AI models.

Together, these advances demonstrate how automated experimentation, intelligent optimization, and molecular-scale insight can accelerate the development of next-generation cooling fluids for electrification, advanced computing, and other energy-intensive applications.